Abstract

A constraint relaxation based multilevel optimization approach was proposed recently for the optimization of large engineering systems, in which the original system was decomposed into a number of smaller subsystems and a coordinate system. The optimization of subsystems can be executed concurrently so as to shorten the computation time. In this paper, parallelization of the approach will be implemented on an Origin 2000 supercomputer, by distributing one subsystem into one CPU. The parallel efficiency and the speedup performance will be investigated and compared with those obtained on a network computing environment.